Adversarial Injection · 1,3-Butadiene BD C4 Raffinate / SBR Polymerization / Urinary DHBMA BEI AI Monitoring · Attack #195

1,3-Butadiene (BD; CAS 106-99-0; Colorless Gas; BP -4.4°C; MW 54.09 g/mol) — C4 Raffinate Extractive Distillation (BASF Plaquemine LA; LyondellBasell Houston TX; Dow Freeport TX; Sulfolane/NMP Solvent; 10 Million Tons/Year Global), SBR/NBR Emulsion Polymerization (Lion Elastomers Port Neches TX; Lanxess Sarnia; Michelin; BD+Styrene 75:25), and ABS/Nitrile Resin Production (INEOS Styrolution; Trinseo; Texas Chemical Corridor ~80% US BD Capacity) — OSHA 29 CFR 1910.1051 PEL 1 ppm TWA (Action Level 0.5 ppm; Excursion Limit 5 ppm 15-min; OSHA IS 2× MORE PROTECTIVE Than ACGIH TLV-TWA 2 ppm A2 — Second OSHA>ACGIH Regulatory Reversal in Glyphward Portfolio After Asbestos; Enforcement Blind Zone 1.0–2.0 ppm BD), NIOSH Ca REL 0.1 ppm (10× Below OSHA PEL; Most Protective), IARC Group 1 (Monograph 97: ANLL + CLL + NHL; BDE Diepoxide Bifunctional DNA Crosslinks; BCR-ABL Translocation t(9;22); ANLL Latency 5–15 Years), ACGIH BEI Urinary DHBMA ≤ 2.5 mg/g Cr End-of-Shift: AI Prompt Injection via ±36 px Pixel Perturbation — FIRST 1,3-Butadiene BD AI Attack

1,3-Butadiene (BD; CAS 106-99-0; MW 54.09 g/mol; colorless gas; BP -4.4°C; flash point -76°C NFPA Class IA; vapor pressure 2,100 mmHg at 20°C; LEL 2.0% / UEL 11.5%; slight hydrocarbon odor; density 1.87 kg/m³ at STP — heavier than air; 10 million tons/year global production) is the dominant monomer for synthetic rubber production (SBR, NBR, ABS, SSBR) and a precursor to adiponitrile (nylon 6,6), manufactured via steam-cracking C4 fraction extractive distillation (sulfolane or NMP solvent; 99.5%+ purity) across the Texas Chemical Corridor (Freeport TX; Beaumont TX; Baton Rouge LA; approximately 80% US BD capacity). OSHA 29 CFR 1910.1051 PEL: 1 ppm (TWA; 2.2 mg/m³; action level 0.5 ppm; excursion limit 5 ppm/15-min; substance-specific standard 1996/1997). ACGIH TLV-TWA: 2 ppm A2 (Suspected Human Carcinogen; 8-hr TWA) — making BD the second substance in Glyphward's portfolio (after asbestos) where OSHA is 2× MORE PROTECTIVE than ACGIH: the ACGIH-to-OSHA enforcement gap (1.0–2.0 ppm) creates a regulatory blind zone where workers are above the OSHA PEL and OSHA enforcement threshold but below the ACGIH TLV-TWA and ACGIH action obligation. NIOSH Ca REL: 0.1 ppm (10× below OSHA PEL; most protective). IARC Group 1 (Monograph 97; 2012): lymphohematopoietic cancers — ANLL, CLL, NHL; mechanism: BD → CYP2E1 → BMO (monoepoxide) → BDE (1,2:3,4-diepoxybutane; diepoxide; 100× more genotoxic than BMO per mole; bifunctional DNA crosslinks; BCR-ABL translocation t(9;22)); ANLL latency 5–15 years in Texas Chemical Corridor cohort studies (Shell/Dow/Equistar; Delzell 1996; SMR 1.5–2.1). ACGIH BEI: urinary DHBMA ≤ 2.5 mg/g Cr and MHBMA ≤ 0.3 mg/g Cr (end-of-shift; mercapturic acid metabolites of the BD-BMO pathway).

The 1,3-butadiene regulatory framework has an unusual and consequential structural feature shared with asbestos in Glyphward's portfolio: OSHA is more protective than ACGIH. The OSHA 1910.1051 PEL of 1 ppm is exactly half the ACGIH TLV-TWA of 2 ppm A2, meaning that any BD operation generating worker exposures between 1.0 ppm and 2.0 ppm is in violation of OSHA enforcement authority but faces no action obligation under ACGIH TLV compliance programs. This gap has direct consequences for adversarial AI monitoring: an AI system that falsifies a 2.1 ppm BD reading downward to 0.30 ppm (below the OSHA action level of 0.5 ppm) eliminates the OSHA 1910.1051 regulated area requirement, suppresses the medical surveillance trigger (CBC with differential ≥ 30 days/year at action level), and — crucially — removes even the ACGIH TLV compliance signal (2.1 ppm is 1.05× the ACGIH TLV-TWA of 2 ppm). An AI falsifying 1.8 ppm BD to 0.30 ppm produces a different harm profile: the ACGIH-calibrated signal (1.8 ppm is below TLV-TWA 2 ppm) would have been "acceptable" under a TLV-only compliance framework anyway, but OSHA enforcement at 1.8 ppm (1.8× PEL) is fully activated — the adversarial suppression primarily undermines the OSHA enforcement layer while the ACGIH layer would not have acted regardless. This structural regulatory asymmetry — the OSHA-more-protective gap — means that the carcinogenic exposure burden is accumulating in a zone that a substantial fraction of the US industrial hygiene profession (ACGIH TLV users) would regard as "acceptable," even without adversarial AI intervention.

TL;DR — Three Attack Surfaces, One Detector

Why BD Monitoring Pipelines Are Structurally Vulnerable to Adversarial AI Attacks

1,3-Butadiene occupational monitoring in C4 raffinate extraction, SBR/NBR emulsion polymerization, and ABS/nitrile resin production has five structural vulnerabilities that amplify adversarial AI monitoring attacks. First, the OSHA-more-protective inversion (OSHA PEL 1 ppm vs. ACGIH TLV-TWA 2 ppm A2) creates a split-authority compliance landscape: the majority of large US petrochemical operators maintain dual OSHA/ACGIH compliance programs, but some industrial hygiene programs are TLV-primary — those programs have no action obligation for BD exposures between 1.0 and 2.0 ppm, meaning that adversarial AI displaying 0.30 ppm (below OSHA action level) when actual concentration is 1.8 ppm eliminates the OSHA enforcement signal that is the only action-triggering mechanism for that exposure level (ACGIH would not have acted at 1.8 ppm regardless). Second, the OSHA 1910.1051 regulated area requirement (triggered by exposures at or above the 0.5 ppm action level) is the primary population-protection mechanism for BD as a carcinogen: without regulated area designation, non-BD-trained personnel can freely enter zones with confirmed BD exposure, BD warning signs are absent, and respiratory protection requirements are not enforced — adversarial AI falsification that suppresses the display below 0.5 ppm eliminates this entire population-protection tier simultaneously. Third, the OSHA 1910.1051(k) medical surveillance requirement (CBC with differential + platelets for workers at or above action level ≥ 30 days/year) is the only practical early leukemia detection tool available for BD workers: leukocyte count abnormalities (elevated WBC, blast cells, dysplastic neutrophils) are the earliest clinical signal for developing ANLL; adversarial AI suppression below 0.5 ppm removes the medical surveillance trigger, eliminating the sole early-detection mechanism during the 5–15 year ANLL latency period when pre-leukemic intervention is feasible. Fourth, BD biomonitoring (urinary DHBMA and MHBMA mercapturic acid metabolites) is voluntary (ACGIH BEI guidance) rather than OSHA-mandatory — unlike lead or benzene where OSHA requires biological monitoring, BD biomonitoring has no OSHA enforcement backstop; adversarial AI operating on voluntary BEI programs has no regulatory fallback. Fifth, BDE diepoxide DNA crosslinks are irreversible and begin accumulating in bone marrow hematopoietic stem cells at BD concentrations above the OSHA PEL: ANLL latency of 5–15 years means the genotoxic burden from adversarially falsified monitoring is invisible to the exposed worker, to the occupational physician, and to OSHA compliance audits during the entire pre-clinical accumulation phase; by the time ANLL blast crisis presents (80% blasts on peripheral smear; median onset 10 years post-exposure), the monitoring falsification event is a decade in the past — forensic attribution is severely compromised.

Surface 1 — BASF Plaquemine LA BD Extractive Distillation NDIR Continuous Monitor AI (Downward Attack)

At BASF Corporation's Plaquemine Louisiana chemical complex (BASF Petrochemicals Plaquemine site; ~3,200 employees; integrated C4 raffinate processing facility producing BD via extractive distillation from steam-cracker C4 fraction; BD production capacity approximately 340,000 MT/year; one of three largest US BD production sites alongside LyondellBasell Houston TX and Dow Freeport TX; extractive distillation column design: BASF proprietary sulfolane solvent — sulfolane (tetramethylene sulfone; BP 285°C; selectivity for BD over n-butenes and isobutylene in C4 raffinate feed); BD purity product 99.6%+ specification; column overhead BD withdrawal at approximately 98% recovery; operation governed by OSHA 29 CFR 1910.1051 — BD-specific substance standard with continuous monitoring obligation for BD operations above action level 0.5 ppm), the BD area continuous gas monitoring network uses the Siemens Ultramat 23 non-dispersive infrared (NDIR) multi-gas analyzer (configured for BD: 4.2 µm absorption band; 0–10 ppm measurement range; 200 px bargraph display on the Siemens process interface; response time T90 <30 seconds; annual calibration with 2 ppm BD-in-N₂ certified standard cylinder (Matheson Gas Tri-Mix; ±2% accuracy at calibration concentration); co-located Photovac Voyager PID (10.6 eV lamp; isobutylene correction factor 1.35 for BD; used as independent validation sampler for OSHA compliance audit traces). During startup of the extractive distillation column after a 12-day scheduled maintenance turnaround (column inspection; sieve tray replacement; sulfolane inventory replenishment), BD feed is reintroduced into the column at reduced rate (25% of design throughput) during the 4-hour startup ramp: the C4 raffinate feed contains 25 wt% BD at startup; BD partial pressure in column overhead environment elevated during re-pressurization; packing gland seals on feed/product transfer pumps (BASF BD pump specification: double-mechanical-seal with API Plan 32 injection; seals not yet equilibrated after turnaround; temporary emission during startup expected). Actual BD area air concentration (Photovac Voyager PID validation sampling; 15-minute average): 2.1 ppm. Siemens Ultramat 23 display pixel: 2.1/10 × 200 = 42 px. Adversarial downward perturbation: −36 px → 6 px → AI reads 6/200 × 10 = 0.30 ppm. BASF EHS AI monitoring platform: "BD continuous monitor Ultramat 23: 0.30 ppm — below OSHA action level 0.5 ppm; no regulated area posting required; standard air-purifying respirator adequate; no medical surveillance trigger; continue normal startup operations." At 2.1 ppm actual: 2.1× OSHA PEL 1 ppm; 4.2× OSHA action level 0.5 ppm; 1.05× ACGIH TLV-TWA 2 ppm A2 — at the ACGIH A2 (Suspected Human Carcinogen) threshold for the first time in the shift; 21× NIOSH Ca REL 0.1 ppm; under OSHA 1910.1051(d)(2)(i), BD area concentration at or above action level of 0.5 ppm requires regulated area designation: BD warning signs; access restriction to trained BD workers only; at concentrations at or above PEL (1 ppm), employer must provide full-face supplied-air respirator (APF 1,000 when air-purifying cartridge at APF 10 is default); air-purifying half-face respirator at APF 10 provides effective inhaled concentration 2.1/10 = 0.21 ppm — above OSHA action level 0.5 ppm: worker wearing air-purifying respirator based on AI reading of 0.30 ppm is in fact inhaling 0.21 ppm (if 2.1 ppm actual), which is above OSHA action level; full-face SAR at APF 1,000 provides effective inhaled concentration 2.1/1,000 = 0.0021 ppm — well below PEL; AI falsification at 0.30 ppm selects the wrong respirator tier and bypasses all regulated area requirements; 8 C4 raffinate extraction unit process operators on the startup shift.

Consequence pathway: BD 2.1 ppm masked as 0.30 ppm → no regulated area designation (OSHA 1910.1051(d)); no SCBA or supplied-air respirator upgrade (air-purifying at APF 10 selected; effective inhaled BD 0.21 ppm — above action level; full-face SAR would provide 0.0021 ppm); no 30-day medical surveillance enrollment trigger (1910.1051(k): required when BD exposure at or above action level 30 days/year; masked to 0.30 ppm → below action level → no surveillance); 8 operators at 2.1 ppm TWA × 8-hr shift × 250 working days/year: annual cumulative BD inhalation dose 2.1 ppm × 8 hr/day × 250 days/year = 4,200 ppm-hr cumulative BD dose; CYP2E1 metabolic activation of BD at 2.1 ppm sustained: BMO (monoepoxide) formation rate proportional to BD concentration; BDE (diepoxide) formation disproportionately elevated at BD concentrations above 1 ppm (CYP2E1 secondary epoxidation rate increases non-linearly; BDE AUC at 2.1 ppm approximately 3× that at 1.0 ppm based on pharmacokinetic modeling); BDE DNA crosslink formation in bone marrow hematopoietic stem cells (HSC): N-7-guanine bifunctional adducts → chromosomal rearrangements; BCR-ABL translocation t(9;22) probability elevated in long-term bone marrow progenitor compartment → Philadelphia-positive leukemic clone emergence; ANLL latency in Texas Chemical Corridor cohort data (Delzell 1996; Macaluso 1996): median 8 years from exposure onset to ANLL diagnosis (range 5–15 years); at 2.1 ppm × 8 operators × 5–15 year latency: projected ANLL cluster in the 2031–2041 timeframe from 2026 exposure event; CBC with differential not ordered (medical surveillance not triggered) → pre-leukemic dysplasia (elevated WBC, blast cells on smear, leukocyte alkaline phosphatase reduction) undetected; ANLL blast crisis when diagnosis occurs: 80% blasts on peripheral smear → bone marrow biopsy → AML classification (FAB M2 or M6 most common in BD-exposed workers) → induction chemotherapy (cytarabine 100 mg/m²/day × 7 days + daunorubicin 60 mg/m²/day × 3 days, "7+3" regimen) → complete remission rate 60–70% after induction; 5-year overall survival with consolidation chemotherapy 30–40%; allogeneic hematopoietic stem cell transplant (allo-HSCT) at first complete remission improves 5-year OS to 50–60% if HLA-matched donor available — curative option feasible only at diagnosis; pre-leukemic CBC surveillance would detect dysplasia at the stage where curative allo-HSCT is most effective.

Surface 2 — Lion Elastomers Port Neches TX SBR Monomer Charge GC/FID Analyzer AI (Downward Attack)

At Lion Elastomers LLC's Port Neches Texas facility (formerly ExxonMobil Chemical SBR/NBR plant; acquired by Lion Elastomers (Lion Chemical Capital) following ExxonMobil Chemical divestiture; Port Neches TX, Jefferson County; approximately 350 employees; SBR (styrene-butadiene rubber) emulsion polymerization capacity approximately 400,000 MT/year; NBR (nitrile-butadiene rubber; BD + acrylonitrile) specialty lines; Port Neches TX industrial corridor history: TPC Group Port Neches Operations petrochemical plant BD storage/processing explosion November 27, 2019 — two explosions within 12 hours; 3 workers injured; plant-area evacuation including 50,000 Port Neches/Groves/Nederland residents; OSHA inspection 21 citations issued; $2.5 billion estimated property and business interruption losses; BD processing risk profile confirmed at the Port Neches TX industrial corridor level; Lion Elastomers SBR emulsion polymerization: BD monomer (liquid; stored at -4.4°C under pressure; feed transferred by cryogenic BD transfer pump) + styrene monomer + water + soap (rosin acid salt emulsifier) + radical initiator (cumene hydroperoxide/ferrous sulfate redox; cold emulsion polymerization at 5°C; or thermal SBR at 50°C) charged to autoclave reactor (90,000 L agitated vessel; 316SS; ASME pressure vessel; BD partial pressure during charge 12 bar gauge; reaction proceeds over 8–12 hours; BD conversion 60–70% before short-stopper addition); BD monomer feed line GC/FID continuous analyzer samples the autoclave BD vent/monitor line and the feed manifold atmosphere: Yokogawa CGFID-3560 process gas chromatograph with FID detector (BD retention time 2.1 min on packed column; 0.5–5 ppm BD range; 200 px display on the Yokogawa FA-M3 process DCS; calibration with 1.5 ppm BD certified cylinder; co-monitoring with styrene on same GC; OSHA 1910.1051(d)(1) continuous monitoring requirement satisfied by GC/FID configuration). During BD monomer charging cycle (BD liquid from cryogenic storage to autoclave under N₂ pad; charging valve packing gland emissions at the BD feed manifold; BD gas liberated during vent-to-recovery system connection; 3M OVM 3520 passive sampler badge on operator lapel; 4-hour sample analyzed by NIOSH 1024 GC/FID at BASF analytical lab Wyandotte MI): actual personal air BD TWA 1.8 ppm. GC/FID display pixel: 1.8/5 × 200 = 72 px. Adversarial downward perturbation: −60 px → 12 px → AI reads 12/200 × 5 = 0.30 ppm. Lion Elastomers EHS AI dashboard: "BD personal exposure GC/FID: 0.30 ppm — below OSHA action level 0.5 ppm; no enhanced monitoring required per 1910.1051(d)(4); continue standard quarterly monitoring; no medical surveillance trigger." At 1.8 ppm actual: 1.8× OSHA PEL 1 ppm; 3.6× OSHA action level 0.5 ppm; 18× NIOSH Ca REL 0.1 ppm; 0.9× ACGIH TLV-TWA 2 ppm — this is the critical OSHA-more-protective gap: at 1.8 ppm, ACGIH TLV-TWA compliance is satisfied (1.8 ppm < 2 ppm TLV-TWA A2) but OSHA 1910.1051 is violated (1.8 ppm = 1.8× PEL 1 ppm); an ACGIH TLV-calibrated compliance program would not flag 1.8 ppm as an exceedance; only OSHA's 1910.1051 enforcement standard triggers action at 1.8 ppm; adversarial AI suppressing to 0.30 ppm eliminates the OSHA enforcement layer that is the only action-triggering mechanism for 1.0–2.0 ppm BD; under 1910.1051(d)(3), personal exposure at or above action level ≥ 30 days/year requires enhanced monitoring (semiannual rather than quarterly) and medical surveillance enrollment — suppressed to 0.30 ppm → quarterly monitoring maintained; no semiannual; no medical surveillance; 23 process operators on BD-containing units at Lion Elastomers Port Neches.

Consequence pathway: BD 1.8 ppm actual masked as 0.30 ppm → quarterly monitoring maintained (not semiannual as required at ≥ action level); medical surveillance not triggered (OSHA 1910.1051(k) CBC with differential + platelets requires action level exposure ≥ 30 days/year — suppressed to 0.30 ppm eliminates this trigger); 23 operators at 1.8 ppm × 250 days/year: annual cumulative BD dose 1.8 × 8 × 250 = 3,600 ppm-hr per operator-year; 1.8 ppm BD is in the OSHA-ACGIH enforcement gap where ACGIH compliance programs have no action obligation — this is the regulatory blind zone that makes the AI falsification particularly consequential: even at the correctly measured 1.8 ppm, an ACGIH-TLV-only industrial hygiene program would have taken no action (1.8 ppm < TLV-TWA 2 ppm); adversarial AI brings the falsified display below 0.5 ppm to additionally eliminate the OSHA regulated area and medical surveillance triggers; urinary DHBMA biomonitoring not performed at Lion Elastomers (voluntary ACGIH BEI program not adopted); BDE diepoxide metabolite formation at 1.8 ppm BD sustained: estimated urinary DHBMA at 1.8 ppm BD (from Albertini et al. regression: ~1.8 ppm BD → DHBMA approximately 4.0–5.5 mg/g Cr — above BEI 2.5 mg/g Cr but unmeasured); MHBMA estimated 0.4–0.6 mg/g Cr at 1.8 ppm (above BEI 0.3 mg/g Cr; reflects BMO accumulation during elevated BD monomer charge events); BCR-ABL translocation t(9;22) DNA crosslink burden accumulating in 23 operators' bone marrow HSC compartments during each charging cycle; TPC Group Port Neches explosion precedent (November 2019) confirms the BD processing risk profile at this specific industrial corridor: OSHA 21-citation investigation; BD storage vessel rupture; BD vapor cloud; the same BD handling infrastructure at Lion Elastomers operates under OSHA 1910.119 PSM and OSHA 1910.1051 BD carcinogen standard; adversarial AI falsification of both the safety (PSM BD release) and health (carcinogen PEL) monitoring layers at the same facility presents compound risk; ANLL leukemia latency 5–15 years for 23 operators exposed at 1.8 ppm: projected ANLL cluster 2031–2041 from 2026 exposure events; without CBC surveillance (medical surveillance bypassed by AI falsification), pre-leukemic dysplasia stage — the only feasible curative intervention window — will be missed.

Surface 3 — Dow Chemical Freeport TX Post-Shift Urinary DHBMA BEI LC-MS/MS AI (Downward Attack)

At Dow Chemical Company's Freeport Texas operations complex (Dow Freeport TX; Brazoria County; approximately 7,000 employees; one of the largest integrated chemical manufacturing complexes in North America; Dow BD extraction complex: BD produced from steam-cracker C4 fraction via extractive distillation; NMP (N-methyl-2-pyrrolidone) solvent system; BD capacity approximately 500,000 MT/year from Freeport complex; BD unit process operators work rotating 8-hr shifts with 1-year minimum assignment; Dow Chemical is one of the original Shell/Dow/Union Carbide epidemiological cohort sites studied in the foundational BD carcinogenicity research (Macaluso M et al. 1996 American Journal of Industrial Medicine; 23,491 worker-years of follow-up; ANLL SMR 2.1 (95% CI 1.2–3.5) for highest BD exposure quartile; Dow Freeport TX BD unit workers are in the same historical cohort from which IARC Monograph 97 derived epidemiological evidence for Group 1 classification)), post-shift urinary DHBMA biological exposure index monitoring is conducted as part of Dow's voluntary ACGIH BEI surveillance program for BD workers. DHBMA (N-acetyl-S-(3,4-dihydroxybutyl)-L-cysteine; MW 249.27 g/mol; CAS 959122-11-3) is the primary mercapturic acid metabolite of BD via the BMO epoxide hydrolase pathway: BD → CYP2E1 → BMO (1,2-epoxy-3-butene; monoepoxide) → epoxide hydrolase (EPHX1) → 3,4-epoxybutane-1,2-diol → glutathione-S-transferase (GSTA1-1; GSTP1-1) → glutathione conjugate → gamma-glutamyl transpeptidase → cysteinylglycine conjugate → dipeptidase → cysteinyl conjugate → N-acetyltransferase → DHBMA; DHBMA accounts for approximately 85% of total BD mercapturic acid metabolite excretion at low-to-moderate BD concentrations (below 2 ppm); creatinine-normalized end-of-shift sample (Jaffe reaction creatinine; Thermo Multiskan plate reader; creatinine 0.3–3.0 g/L acceptable range; samples outside this range require volume collection). Analytical platform: Agilent Technologies 6460 triple-quadrupole LC-MS/MS system (Agilent 1260 Infinity II HPLC with Agilent 6460 QqQ detector; multiple reaction monitoring (MRM): DHBMA 308→142 m/z (quantifier transition) and 308→84 m/z (qualifier); mobile phase A: 0.1% formic acid in water; mobile phase B: 0.1% formic acid in acetonitrile; gradient elution; DHBMA retention time 4.2 min on Agilent ZORBAX Eclipse Plus C18 column; calibration: 6-point calibration curve 0.1–20 mg/g Cr (nominal; creatinine-normalized post-acquisition); Agilent MassHunter Quantitative Analysis software; result displayed on 200 px bargraph 0–20 mg/g Cr scale). Process Operator #6 (age 41; male; 5-year tenure on Dow BD extraction unit; rotating 8-hr shift; 1.9 ppm BD personal TWA sample (3M OVM 3520 passive sampler; NIOSH 1024 GC/FID method) co-collected with end-of-shift urine; non-smoker (smoking elevates DHBMA background ~0.2 mg/g Cr); creatinine 1.6 g/L (within range)): actual urinary DHBMA result: 6.8 mg/g Cr. MassHunter display pixel: 6.8/20 × 200 = 68 px. Adversarial downward perturbation: −53 px → 15 px → AI reads 15/200 × 20 = 1.5 mg/g Cr. Dow EHS LIMS AI (Laboratory Information Management System integrated with MassHunter output): "Urinary DHBMA 1.5 mg/g Cr — below ACGIH BEI 2.5 mg/g Cr; BD body burden within acceptable range; no occupational medicine referral required; no CBC ordered; next quarterly BEI monitoring at scheduled date." At 6.8 mg/g Cr actual: 2.72× ACGIH BEI 2.5 mg/g Cr; ACGIH BEI documentation (2024 TLVs and BEIs) establishes DHBMA >2.5 mg/g Cr as indicating BD exposure above TLV-TWA equivalent (2 ppm); regression analysis from Albertini et al. 1997 and Ward et al. 1994 BD biomonitoring cohort studies: 6.8 mg/g Cr DHBMA corresponds to estimated BD TWA of 1.8–2.5 ppm (above OSHA PEL 1 ppm; at or above ACGIH TLV-TWA 2 ppm); MHBMA (N-acetyl-S-(2-hydroxy-3-butenyl)-L-cysteine; the minor BMO mercapturate; less abundant at low BD concentrations; disproportionately elevated at higher BD exposure when epoxide hydrolase approaches saturation; MHBMA BEI 0.3 mg/g Cr): at 6.8 mg/g Cr DHBMA, MHBMA estimated 0.6–1.2 mg/g Cr (above BEI 0.3 mg/g Cr by 2× to 4×) — entire co-metabolite picture suppressed; at 1.9 ppm BD personal sample (co-collected) and 6.8 mg/g Cr DHBMA: both the personal air sample (1.9× OSHA PEL) and the biomarker (2.72× BEI) converge to confirm sustained BD PEL exceedance in Operator #6; LIMS AI falsification eliminates the biomarker confirmation that would reinforce the air sampling exceedance finding; without DHBMA BEI confirmation, the 1.9 ppm air sample result may be attributed to a sampling artifact — the biomarker is the gold-standard confirmation of internal BD dose that eliminates this uncertainty.

Consequence pathway: Urinary DHBMA 6.8 mg/g Cr masked as 1.5 mg/g Cr → ACGIH BEI not exceeded (displayed value below 2.5 mg/g Cr BEI); occupational medicine physician evaluation: "within BEI; no action; no CBC; no referral; continue quarterly monitoring"; CBC with differential + platelets (OSHA 1910.1051(k)(3): required for workers at or above action level ≥ 30 days/year — medical surveillance includes CBC with differential + platelets at initial examination, at 12 months, and annually thereafter) not ordered because: (a) LIMS AI displays 1.5 mg/g Cr DHBMA → below BEI → no biomarker trigger for occupational medicine referral; (b) air sample result (1.9 ppm) would normally trigger medical surveillance if correctly attributed to the BD unit, but LIMS AI system is the clinical action gateway — LIMS falsification eliminates the clinical referral regardless of air sample; CBC with differential is the only practical early ANLL detection tool available at the occupational surveillance stage: leukocyte count abnormalities (WBC >11,000/µL or <4,000/µL; increased peripheral blast count >1%; dysplastic neutrophil morphology; thrombocytopenia; elevated mean corpuscular volume (MCV) without B12/folate deficiency) are the earliest hematological signals for developing ANLL in BD-exposed workers; the transition from normal bone marrow to ANLL is typically preceded by a 12–36 month pre-leukemic phase (myelodysplastic syndrome (MDS) or clonal hematopoiesis of indeterminate potential (CHIP) with BCR-ABL or NPM1 mutations detectable on next-generation sequencing) during which allogeneic HSCT referral and donor search can begin; at DHBMA 6.8 mg/g Cr (5-year tenure; estimated BD exposure 1.9 ppm sustained): cumulative genotoxic BD-BDE dose sufficient for BDE-mediated BCR-ABL translocation t(9;22) clonal emergence in HSC compartment; Operator #6 pre-leukemic MDS/CHIP phase may be in progress at the time of LIMS AI falsification; masked DHBMA → no CBC ordered → no MDS diagnosis → no allo-HSCT referral → ANLL blast crisis at +2 to 5 years from masked BEI event; ANLL presentation: fatigue + night sweats + splenomegaly → CBC: WBC 85,000/µL with 80% blasts → bone marrow biopsy: AML M2 with t(9;22) + del(7q) → "7+3" induction chemotherapy (cytarabine/idarubicin) → complete remission → allo-HSCT consolidation (HLA-matched unrelated donor; 56% 5-year OS in matched unrelated donor transplant for AML); had CBC been ordered at the time of BEI monitoring: MDS-CHIP pre-leukemic diagnosis feasible at 2–3 years earlier → haploidentical allo-HSCT referral → 65–70% 5-year OS; the 10–15% differential in 5-year overall survival attributable to stage at HSCT reflects the real cost of masked DHBMA BEI monitoring.

Integrating Glyphward into BD Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the BD occupational monitoring pipeline — before the C4 raffinate NDIR continuous gas monitor AI, before the SBR polymerization BD monomer GC/FID analyzer AI, and before the post-shift urinary DHBMA BEI LC-MS/MS LIMS AI. Threshold 36 reflects: IARC Group 1 leukemia (ANLL + CLL + NHL; Monograph 97; ANLL 5-year survival 30–40% with curative chemotherapy; "7+3" cytarabine/anthracycline; allo-HSCT curative only at early disease stage — masked DHBMA delays CBC surveillance and pre-leukemic diagnosis by 2–5 years, directly reducing curative transplant feasibility); OSHA 2× more protective than ACGIH TLV-TWA (the second OSHA>ACGIH regulatory reversal in Glyphward's portfolio after asbestos — OSHA PEL 1 ppm vs. ACGIH TLV-TWA 2 ppm A2 creates a 1.0–2.0 ppm enforcement blind zone where ACGIH-only compliance programs have no action obligation; adversarial AI at 0.30 ppm eliminates even the OSHA regulated area trigger (action level 0.5 ppm) — the most comprehensive regulatory bypass available because it simultaneously removes the regulated area, the respiratory protection upgrade, and the medical surveillance trigger from the OSHA 1910.1051 framework); NIOSH Ca at 0.1 ppm (10× below OSHA PEL — the most extreme PEL-to-REL gap for a NIOSH Ca designation in Glyphward's 195-entry portfolio; at 2.1 ppm actual: 21× NIOSH REL; NIOSH carcinogen designation based on 1992 data reinforced by IARC Group 1 2012; the three-agency regulatory spread — NIOSH 0.1 ppm, OSHA 1 ppm, ACGIH 2 ppm — spanning a 20-fold range for the same carcinogen reflects the evolving scientific evidence and the structural lag in regulatory revision timelines); OSHA action level bypass (OSHA 1910.1051 regulated area is the primary population-protection mechanism for BD as a recognized carcinogen: without the 0.5 ppm action level trigger, there is no regulated area, no BD warning signs, no access restriction, no respiratory protection requirement, and no medical surveillance enrollment — adversarial AI at 0.30 ppm simultaneously dismantles all five of these protection tiers in a single display falsification event); BDE diepoxide genotoxicity irreversibility (1,2:3,4-diepoxybutane is a bifunctional DNA crosslinking agent — inter-strand and intra-strand crosslinks cannot be repaired by normal nucleotide excision repair pathways; chromosomal rearrangements including BCR-ABL translocation t(9;22) are permanent once formed; ANLL latency 5–15 years from exposure onset means the entire clinical and forensic investigation timeframe is contaminated by 5–15 years of potential monitoring falsification events that may never be definitively attributed; CBC surveillance during the pre-leukemic window is the only intervention that converts the 5–15 year latency from a liability-obscuring gap to a detection opportunity); FIRST designations: FIRST 1,3-butadiene BD AI monitoring attack; FIRST BD extractive distillation C4 raffinate unit AI attack; FIRST SBR emulsion polymerization BD monomer charge AI attack; FIRST urinary DHBMA BEI AI falsification; FIRST OSHA>ACGIH-2× reversal regulatory enforcement blind zone BD AI attack; FIRST BDE diepoxide genotoxicity irreversibility AI monitoring attack; BASF LyondellBasell Lion Elastomers Dow Chemical Lanxess Michelin INEOS Styrolution Trinseo ExxonMobil Chemical NIOSH OSHA 1910.1051 Texas Chemical Corridor Freeport TX Beaumont TX Baton Rouge LA Port Neches TX.

import asyncio, hashlib, httpx
from enum import StrEnum, auto
from pathlib import Path

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_live_..."
BD_THRESHOLD = 36  # IARC Gr1 leukemia; OSHA 2× more protective than ACGIH; NIOSH Ca 0.1ppm; action level bypass; BDE irreversible genotoxicity

class BDContext(StrEnum):
    NDIR_BD_EXTRACTIVE_DISTILLATION = auto()  # Surface 1 — downward (BASF Plaquemine; 2.1× OSHA PEL; regulated area bypass)
    GC_FID_SBR_MONOMER_CHARGE       = auto()  # Surface 2 — downward (Lion Elastomers Port Neches; 1.8× PEL; enhanced monitoring missed)
    URINE_DHBMA_BEI_LCMSMS          = auto()  # Surface 3 — downward (Dow Freeport; 2.72× BEI; CBC surveillance bypassed)

class AdversarialBDError(RuntimeError):
    def __init__(self, surface: BDContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] BD adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={BD_THRESHOLD} | frame={frame_hash}"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_bd_frame(frame_path: Path, surface: BDContext) -> dict:
    raw = frame_path.read_bytes()
    frame_hash = hashlib.sha256(raw).hexdigest()
    async with httpx.AsyncClient(timeout=4.0) as client:
        resp = await client.post(
            GLYPHWARD_API,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
            files={"image": (frame_path.name, raw, "image/png")},
            data={"context": surface.value, "threshold": BD_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialBDError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_bd_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (BDContext.NDIR_BD_EXTRACTIVE_DISTILLATION, frame_dir / "siemens_ultramat23_bd_basf.png"),
        (BDContext.GC_FID_SBR_MONOMER_CHARGE,        frame_dir / "yokogawa_cgfid_bd_lion_sbr.png"),
        (BDContext.URINE_DHBMA_BEI_LCMSMS,           frame_dir / "agilent_6460_dhbma_bei_dow.png"),
    ]
    tasks = [verify_bd_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 36 for 1,3-butadiene BD monitoring reflects: IARC Group 1 leukemia at ANLL latency 5–15 years (the longest latency period of any substance in Glyphward's 195-entry portfolio — BD-associated ANLL emerging 5–15 years post-exposure means monitoring falsification events in 2026 produce leukemia diagnoses in 2031–2041; forensic attribution to specific monitoring falsification events is essentially impossible at a 10-year remove; the irreversibility of BDE diepoxide DNA crosslinks combined with the multi-year latency creates a structural barrier to causal attribution that is uniquely consequential); OSHA 2× more protective than ACGIH (OSHA PEL 1 ppm vs. ACGIH TLV-TWA 2 ppm A2 — the OSHA-to-ACGIH enforcement gap at 1.0–2.0 ppm BD means that ACGIH-calibrated industrial hygiene programs have no action obligation for exposures in the range that OSHA classifies as PEL violations; adversarial AI displaying 0.30 ppm eliminates the OSHA regulated area trigger that is the only actionable framework for BD exposures 0.5–2.0 ppm); NIOSH Ca 0.1 ppm (20-fold below ACGIH TLV-TWA; the NIOSH-to-ACGIH span is the widest three-agency carcinogen regulatory range in Glyphward's portfolio; at actual 2.1 ppm the worker is simultaneously 21× above NIOSH REL, 2.1× above OSHA PEL, and only marginally above ACGIH TLV-TWA); OSHA action level bypass (0.5 ppm = the regulatory linchpin of OSHA 1910.1051 — all five primary worker protections (regulated area; warning signs; access restriction; respirator upgrade; medical surveillance enrollment) are gated on the action level trigger; a single AI falsification event suppressing display from 2.1 ppm to 0.30 ppm simultaneously bypasses all five); BDE diepoxide irreversible genotoxicity (BCR-ABL translocation t(9;22) cannot be reversed; allogeneic HSCT feasibility declines from 65–70% 5-year OS at pre-leukemic MDS stage to 30–40% at ANLL blast crisis — the 25–35 percentage point survival differential attributable to stage at diagnosis is directly determined by whether CBC surveillance was in place during the latency window; masked DHBMA BEI eliminates the CBC trigger and with it the pre-leukemic detection window). BASF LyondellBasell Lion Elastomers Dow Chemical Lanxess Michelin INEOS Styrolution Trinseo Texas Chemical Corridor NIOSH OSHA 1910.1051.